Corrigendum to “A purely confirmatory replication study of structural brain-behavior correlations” [Cortex 66 (2015) 115–133]

In our previous study, we reported a purely confirmatory replication study of structural brain-behavior correlations (Boekel et al., 2015). For all but one of the 17 findings under scrutiny, confirmatory Bayesian hypothesis tests indicated evidence in favor of the null hypothesis ranging from anecdotal (Bayes factor < 3) to strong (Bayes factor > 10). In several studies, effect size estimates were substantially lower than in the original studies. We now discovered a mistake in the post-processing pipeline of our diffusion-weighted imaging (DWI) data analyses originally included in this replication study. This led us to recalculate and correct five of the 17 originally reported brain-behavior correlations that were based on DWI data. In short, after reanalyzing the DWI data correctly, the original conclusions for the five corrected analyses did not change. More concretely, we discovered that an extra volume was included in the acquisition protocol which was subsequently incorrectly included in the data analyses. This extra volume was incorporated due to the Philips scanner software version R3. This volume is the average of all the acquired diffusion weighted volumes and was placed at the end of the data file. Such an extra volume can be used to calculate Apparent Diffusion Coefficient (ADC) maps. This extra volume has a b-value of 1000 and bvecs values of 0,0,0. As this is not truly a measured direction or a proper B0 volume, this volume should have been removed. The extra volume, as well as the corresponding extra entries in the bval and bvecs were removed. All DWI data processing was redone with the pre-registered parameter settings. Removing this extra volume from the analyses resulted in considerably different structural DWI measures including fractional anisotropy (FA), mean diffusivity (MD), and λ1 values from the pre-defined regions of interest (ROIs). This mistake also affected tractography results including the calculation of tract strength. Therefore, the previously reported results regarding the failed replications of Forstmann et al. (2010) and Xu et al. (2012) needed to be corrected. After removing the additional volume from the current DWI data set, the analyses pipeline described in the original paper, i.e., section 1.1.1 DWI analyses and 1.1.2. Probabilistic tractography were used. In addition to the mistake in the post-processing pipeline of the DWI data, it came to our attention that the correlation coefficients reported in the text inset of figure 6 were swapped between the two panels. Although this had no influence on the conclusion, we have taken the opportunity to correct this error. In the following we now present the corrected results, tables, and figures of the studies previously reported (Boekel et al., 2015). 3.1. Replication 1: Forstmann et al. (2010) Summary statistics One subject was removed from the analyses (>2.5 SD from the mean). Below are the corrected new summary statistics for the tract strength measure between the right pre-SMA and right striatum as well as the LBA flexibility measures presented. These are based on 32 subjects and are not corrected for age and gender.[figure presented]